Iterating over the faces of a component - without exploding

talig · 14 Mar 2009, 18:05

I've been trying to iterate over the faces of a component without
exploding it, in the following manner:

Let c be that component.
Then d = c.definitions.entities
for each entity in d - put in in an entities_array.
iterate over the entities_array: If an entity is a group, take it's
entities and add them to an entities_array
if it's a face, add to a result array.

I'm comparing the results of this to that of "Explode".
I get the same number of faces, but the calculations I'm using them
for later disagree.
I was thinking it might be related to the face's Normal Vector, but
that seems improbable to me that using the aforementioned method and
"Explode" will differ in such a manner.

Essentially, the calculation performed is "raytracing" - I'm checking
if a point in a face is visible from another point in the model.

Has anyone seen something like this? Tell me where's the error?
Thanks

Tali

thomthom · 14 Mar 2009, 18:59

hm... not sure. can you post the iterating code snippet?

talig · 21 Mar 2009, 15:48

Certainly. The point is to get an array of faces in the end of it all. I don't need the edges, and I don't count them even when I explode the component in advance.

def get_faces(object)
	entities = []
	type = object.typename
	if type == "Face"
		return [object]
	elsif type == "Loop"
		return [object.face]
	elsif type == "Group"
		entities = to_array(object.entities)
	elsif type == "ComponentInstance"
		entities = to_array(object.definition.entities)
	end
	result = []
	i = 0
	while i<entities.length do
		entity = entities[i]
		type = entity.typename
		if type == "Face"
			result << entity
		elsif type == "Loop"
			result << entity.face
		elsif type == "Group"
			entities.concat(to_array(entity.entities))
		elsif type == "ComponentInstance"
			entities.concat(to_array(entity.definition.entities))
		end
		i+=1
	end
	return result
end

def to_array(entities)
	result = []
	entities.each { |entity| result << entity }
	return result
end

def get_all_selected_faces(selection)
	result = []
	for selected in selection do
		entities = get_faces(selected)
		result.concat(entities)
	end
	return result
end

Jim · 21 Mar 2009, 17:36

Did you know about the raytest method? It sounds like it could help you do a ray trace for visibility.

talig · 21 Mar 2009, 19:27

Jim - Thanks.
I'm using raytest. But I need a fine-grain detail.
See, if I have a component - raytest will return the component as an answer for the ray hitting any of it's faces. I need to know which face in that component was intersected. More accurately, I need to know that a certain face that should have been intersected, isn't - because it's shaded.

Thomthom - any thoughts?

Thanks again guys,

Tali

Chris Fullmer · 21 Mar 2009, 21:17

It also returns the 3dPoint that it hits. You could run that point against all planes of all the faces faces in the component with the .on_plane? method. If it truly hits something, there should be one plane in that compnent that would return true in that comparison.....right?

Chris

talig · 21 Mar 2009, 21:22

Chris -
Thanks again. You are correct, but again - that is not my use case.

I repeat the question:
This code SHOULD work, it's unclear why whatever calculation I'm running on it, disagrees with the result on the exploded component.
Essentially - they both provide an array of faces, and it should be the same faces in both cases, but it seems like it isn't.
Why?

fredo6 · 21 Mar 2009, 21:52

Talig

A few remarks on your script.

You must accumulate the transformations to get the real geometric coordinates of the elements you traverse. If your raytest hits a component 'cp3', embedded in 'cp2' and 'cp1', then you should use


t = cp1.transformation * cp2.transformation * cp3.transformation

Then, you do as Chris suggests, cycle through the faces of component3, and perform 2 tests with the point 'ptray' returned by model.raytest:

Check if on plane (fast test)
check within the face (more greedy, especially when face has holes)


lfaces = []
cp3.definition.entities.each do |e|
    if e.class == Sketchup;;Face
       plane = [t * e.center, t * e.normal]
       lfaces.push e if ptray.on_plane?(plane)
    end
end

myface = nil
tinv = t.inverse
ptlocal = tinv * ptray
lfaces.each do |face|
   status = face.classify_point ptlocal
   if status > 0 && status < 8
      myface = face
      break
   end
end
#myface should contain your face

#Note that if cp3 is a group, you should use cp3.entities instead of cp3.definition.entities.

The list [cp3, cp2, cp1] is returned by the raytestmethod.

As a side remark, you can get an array of entities by using the standard Ruby method: entities.to_areturns a list of entities (this also works on a selection).

Fredo

Chris Fullmer · 21 Mar 2009, 21:58

How large of a component does it take for this to break? I just tried it on a simple group of 5 faces and it returned an array 5 of faces. Are you experiencing problems only on large, multiple nested components?

Chris

fredo6 · 21 Mar 2009, 22:12

I did not test the code (just wrote it on the fly).
The question is whether Raytest returns the point hit in absolute model coordinates, or in local coordinates of the component or group hit. The doc is unclear. In either case, I wanted just to tell Talig that he should use Transformations to cycle through the model.

Fredo

talig · 22 Mar 2009, 11:56

Chris - that's what I said: I get the correctnumber of faces. But I make some calculation about them, and I get a different result when I use explode - though I should get the same one. Even a plain box is not working properly.

Fredo - I'm a female
About the coordinates: What you're saying is interesting. I assumed I get the absolute model coordinates in any case. That's obviously true in the exploded case, but possibly the cause of the problem in what I'm trying to do. I'll check it out. Thanks! (and thanks for the to_a tip!)

I already have everything else up and running, so thanks for trying to help - but really, no call for that.
All I need is an array of faces equivalent to that of an exploded component. Nothing more, nothing less.

Chris Fullmer · 22 Mar 2009, 15:56

@talig said:

Chris - that's what I said: I get the correctnumber of faces. But I make some calculation about them, and I get a different result when I use explode - though I should get the same one. Even a plain box is not working properly.

Heehe, yeah, I noticed that after I was re-reading some of Fredo's posts. I went back and re-read your post and realized that you already were getting the number of faces correct. My mistake.

@talig said:

All I need is an array of faces equivalent to that of an exploded component. Nothing more, nothing less.

I see what your saying now. Makes sense.

So what is the calculations you're trying to run on the faces? Or maybe you've already got it working with Fredo's great explanation?

Chris

thomthom · 22 Mar 2009, 16:08

Have you checked that the list of faces you get is the same as if you exploded them? Not just the number.

talig · 22 Mar 2009, 17:18

thomthom - I don't know how I can check that they are the same.
I mean, I can work something out and save the array before explosion, then explode it and compare, but will the object comparison be good enough? Do I need to compare certain fields? I'll obviously need to sort them both to have them in the same order - which may also be non-trivial...
Ideas?

thomthom · 22 Mar 2009, 21:46

The .to_s method of entities returns something like: #<Sketchup::Face:0xae3ce28>
You could sort the results of each method, exploded and un-exploded, and write it out to two files. When you have your array just use the .sort method. Then you can compare them for differences. For short lists you can do it yourself. For larger lists if you need to test more complex models you can try with some software that compares two files.

Chris Fullmer · 22 Mar 2009, 21:56

I don't think you can .sort a selection set. It always comes back with undefined method '<=>' so I'm guessing they have not built in a way to compare the values of faces to know how to sort them. Even if they implemented sort to work with just the ObjectID string would suffice.

Chris

Chris Fullmer · 22 Mar 2009, 21:57

@chris fullmer said:

Even if they implemented sort to work with just the ObjectID string would suffice.

Oh, that works to do that: selection.to_a.to_s.sort

That will sort it by turning all face ID's into a simple string. Then it can sort them. That will come in handy,

Chris

thomthom · 23 Mar 2009, 07:00

@chris fullmer said:

Oh, that works to do that: selection.to_a.to_s.sort

I was thinking more like selection.to_a.sort
Is there a .sort method for a string?

talig · 23 Mar 2009, 07:41

Thanks thomthom, but who promises that that string is a good identifier, when you run different instances of the program, possibly different files (same model saved under different names), if you have the same component twice in the model? Is it documented anywhere what that hex sequence is? (I'm guessing hash, but even so it matters what it takes into account)

thomthom · 23 Mar 2009, 08:00

You're right. Didn't think of that. The Hash, and .entityID changes, also if you explode the groups.

But, another method: You could make a list of the areas of all faces and compare them. Mind you, if any of your groups/components are scaled you have to take that into account.